From November 2005 through November 2007, a total of 760 cases of human monkeypox in the Sankuru District, DRC, were identified by active disease surveillance and confirmed by quantitative real-time PCR (20). From these, we selected 60 cases on the basis of 3 epidemiologic measures: severity, transmission, and geography. The samples chosen represented the entire set of samples and included samples from every category of severity and transmission observed and every geographic location sampled. Of the 60 cases, we sequenced 29 representative scab or vesicle fluid samples from patients whose samples met the criteria for concentration and quality of genetic material (>500 ng total DNA and >1.8 260/230 and 260/280 absorbance ratio). We used the Genome Analyzer IIx and MiSeq sequencers (Illumina, Inc., San Diego, CA, USA) to conduct 76-bp or 100-bp, paired-end sequencing. To avoid laboratory-induced mutation events, we sequenced these samples directly from scab or vesicle material without virus isolation or propagation.

Techniques for sample collection, processing, nucleotide isolation, and real-time PCR have been described (20,28,29). The World Health Organization scoring system used during the smallpox eradication program was used to generate severity designations: presence of <25 lesions on the body was considered mild disease; 25–99 lesions, moderate; 100–250 lesions, severe, and >250 lesions, serious. Libraries with an average insert size of ≈300 bp were processed by using the TruSeq (Illumina, Inc.) sequencing kit version 5 and sequenced for either 76 or 100 cycles from each end on an Illumina Genome Analyzer IIx or MiSeq instrument.

All raw Illumina sequence data were processed through an in-house–developed filtering program that removes or trims reads with known artifacts. Specifically, all reads containing adaptor sequence, low-complexity reads, or PCR duplicates were removed or trimmed. Host read exclusion was performed by removing reads that aligned to the human genome (Homo sapiens chromosome 10, GRCh37 primary reference assembly) by using the CASAVA pipeline 1.8 (Illumina, Inc.) with default parameters. De novo assemblies were performed by using SeqMan Pro NGen (DNAStar, Inc., Madison, WI, USA) with default parameters. Genome size was set to 200 kb. The optimal k-mer was determined by using an iterative assembly process. We closed gaps in coverage and resolved large repeats by using Sanger sequencing for the 23 genomes described in this work. Sequences were submitted to GenBank with accession nos. JX878407–JX878429.

We conducted the phylogenetic analysis by using methods identical to those reported for VARV diversity (4). In brief, we aligned the CRS (MPXV nucleotide positions ≈56000–120000) of samples with sufficient sequence information with 11 MXPV reference sequences available through the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov/), GenBank nos. KC257459, KC257460, DQ011154, DQ011155, HQ857562, HQ857563, NC_003310, AY603973, AY753185, AY741551, DQ011156, DQ011153, and DQ011157 in SeqMan Pro NGen (DNAStar, Inc.). Manual editing of gaps and ambiguous base calls of the alignment were performed to include the draft genomes and provide a more robust phylogenetic narrative. The resulting alignment was analyzed by using Bayesian methods of phylogenetic reconstruction and the GTR+Γ evolutionary model. MrBayes version 3.1.2 (30) was run by using 1,000,000 generations with 4 parallel runs. Markov chain Monte Carlo convergence was assessed by checking the average standard deviation of split frequencies (<0.01) during >10,000 generations. Figure 1 shows the unrooted tree based on alignment of the 11 MPXV reference sequences and the viral genomes from the 23 clinical samples in this study.

All primers were developed by using Primer-BLAST (www.ncbi.nlm.nih.gov/tools/primer-blast/) with default parameters. This software uses Primer3 to design the primers and then BLAST to determine specificity. We designed qualitative PCRs to detect the left and right short tandem repeats (STRs) at MPXV nucleotide positions 72–486 and 196710–197124, a region spanning the 625-bp deletion between bases 189820 and 190444 in the right ITR, and a 410-bp region of the CRS that contains 4 single-nucleotide polymorphism locations positively typing the lineages described in this article between nucleotide positions 89273 and 89944. All PCRs were run by using standard methods and optimized annealing temperatures of 55°C through 65°C. The PCR product Sanger sequences were aligned to the reference in SeqMan Pro software (DNAStar) for validation by using default parameters. Primer sequences are available on request.

Among the MPXV alignments, we detected a polymorphism in the noncoding region of the ITR with 12 variants. Of the 23 (17.4%) complete genome sequences, 4 showed substantial genomic instability directly upstream of the right ITR. This subset of samples contained a 625-bp deletion between bases 189820 and 190444 (genome positions based on MPXV-COG_2003_358). This deletion completely removes MPV-Z-N2R and eliminates the first 103 bp of the orthopoxvirus major histocompatibility complex class I–like protein (OMCP, MPV-Z-N3R) (Figure 2, panel A). The function of the protein expressed from MPV-Z-N2R is unknown, and no homologous gene is in the VARV or West African MPXV genomes. OMCP is a secreted protein that binds to NKG2D and interferes with NKG2D-mediated cell killing by natural killer cells (31). A conventional PCR created to characterize the deletion enabled us to detect this deletion in 6 additional genomes (Table) We found no evidence of a mixed population within the sequence information for the isolates containing the deletion or in the gel electrophoresis of the Amplicon Sanger sequencing products for samples tested (data not shown).

This genetic polymorphism was absent from all MPXV sequences available through GenBank. Moreover, although 1 full-length copy and 1 truncated copy of the OMCP gene are present in all genomes of the Western African clade MPXV, only the full-length copy of the gene is present in genomes of the Central African MPXV clade (Figure 2, panel A). No homologous gene is present in the VARV genome (26) (www.poxvirus.org). The deletion identified in our clinical samples completely removes the region of OMCP containing the leader sequence (Figure 2, panel B) and substantially alters protein folding (data not shown), suggesting that the truncated protein is nonfunctional. The deletion was identified in genomes belonging to lineage B only (Table). Data regarding route of infection were available for 44 patients; transmission for 24 cases was classified as human-to-human (secondary) and for 20 as animal-to-human (primary) (Table). Metadata were available for 8 of the samples containing the deletion; 7 (87.5%) of the 8 were from cases resulting from human-to-human transmission (p = 0.0544, Fisher exact test), suggesting that this deletion might be associated with increased fitness in humans. Further work, however, is needed to elucidate the role of OMCP in transmissibility.

Substantial variations from other orthopoxviruses were detected in the 70-bp STR region (bases 72–486 and 196710–197124 in MPXV-COG_2003_358) (4,32–34). A traditional PCR designed against the left and right STR regions enabled us to evaluate this polymorphism in all 60 samples. We did not quantitatively test the purity of the repeat population beyond visualization of the PCR product by gel electrophoresis. A small percentage of the samples (10%) had faint secondary products that could not be resolved. The STRs described here should be considered to be the dominant population. In total, we identified 12 unique STR combinations (Table). STR variations have been described for VARV isolates obtained during 1944–1974 (34); however, no studies have addressed the possibility of STR variations in VARV isolates belonging to the same geographic cluster. Samples collected during episodes of VARV activity in Bangladesh, South Africa, or the Far East showed limited variation (<3 left STR patterns); whereas, we observed 12 patterns for MPXV samples collected during a 2-year period in the DRC. Given that the poxvirus genomes in our study were characterized directly from clinical samples, it is plausible that this smaller range of diversity for VARV is only the result of a genetic bottleneck that occurred during cell culture isolation. Rapid copy number variation of the K3L gene in the VACV ITR has recently been described as a mechanism for host immunomodulation (25). However, none of the STR copy number variations observed among the MPXV strains involved a known coding region. In contrast to some VARV isolates, all 60 MPXV samples had identical right and left STRs (Table).

Because only 23 samples contained enough material for complete genome characterization, to expand our analysis to all 60 samples, we performed PCR amplification of a 410-bp variable region of the CRS corresponding to the H4L gene (VACV-COP gene notation). When selecting this region, we used the phylogenetic relationships of the CRS regions of the Central African clade by using 72 segregating sites over 61 kb of sequence. The selected region contained 4 segregating sites that are phylogenetically relevant to Central African clade MPXV (MPXV accession no. DQ01154: 89505, 89545, 89674, and 89914). No evidence of additional diversity was detected at these sites among the samples from the DRC except at the fourth site (89914) and only among the members of lineage B. Phylogenetic reconstruction of this region among the samples resulted in a topologic distribution for DRC isolates that correlates with the whole-genome tree (bootstrap values, as a measure for confidence measures, were lineage A (n = 12, 82%), lineage B (n = 11, 81%), lineage C (n = 6, 80%), and lineage D (n = 1). Moreover, this region contains 7 sites that differentiate between the Central African (n = 29) and Western African (n = 6) clades with a bootstrap value of 82%, which can be evaluated as another measure of the phylogenetically informative character of this genomic area. In this region, we identified 1 nonsynonymous and 3 synonymous changes. An additional nonsynonymous change was identified in all lineages ≈800 bp downstream of this region, for a total of 5 polymorphic sites in H4L within the Central African clade.

Overall, 11.6% of the genetic variations identified in this study were located in an ≈1.2-kb conserved region of the CRS. H4L is reported to be tightly associated with the DNA-dependent RNA polymerase, aiding in early gene transcription before initiation and termination (35,36). Increased mutation rates in a tightly constrained area of a genome can be a marker of increased selective pressure. The effect of the genetic polymorphisms on the function of H4L is unknown, and further studies to evaluate their effects on early gene expression are needed.

The phylogenetically informative single-nucleotide polymorphisms were assessed by PCR amplification and dideoxynucleotide sequencing for all 60 samples. This information was then used to extrapolate the lineage of each sample (Table).

Lineage A contains the widest range of STR repeats (1–15) and represents 33 (55.0%) of the samples analyzed. The lineage seems to have arisen in the Lomela health zone (forest) where it continued to circulate throughout 2007. The remaining members of the lineage are a cluster of strains with 13–15 STR repeats that were geographically restricted to the Vangakete health zone (savannah). According to the World Health Organization scoring system for severity used during smallpox eradication, the virus strains belonging to lineage A caused predominantly moderate disease (66.7%, p = 0.2816). No significant differences in animal-acquired versus human-acquired infection were associated with this lineage.

Lineage B contains 1–8 STRs and represents 16 (26.7%) of the samples analyzed. These strains were predominantly isolated from the Kole (forest) and Lomela health zones and were associated with moderate disease (80.0%, p = 0.2179). Although no statistically significant association with secondary (human) transmission was observed among all lineage B strains, the MPV-Z-N2R/OMCP deletion was identified only in genomes belonging to this lineage and, as stated above, was significantly associated with human-to-human transmission.

Lineage C showed high homogeneity with only 1 STR and represented 10 (6.7%) of the samples analyzed. These strains were predominantly obtained from the Djalo-Ndjeka (ecotone) and Vangakete health zones and were associated with severe/serious disease (70.0%, p = 0.0234).

Only 1 representative of lineage D was identified and contained 10 STRs. The sample was collected in Bene Dibele (forest), was associated with severe disease, and was acquired from a human source.

Although the data do not support causality linking the number of STR repeats to the severity of any lineage, the composition of the STR region merits further investigation for effects on virulence and transmission and as a marker of passaging effects. Our data support a wide diversity of STR composition that varies according to lineage; 1 population exhibited only a dominating single STR repeat and others had variants between 1 and 15 repeats. On the basis of our data and the information previously reported (25) depicting an accordion-like expansion of gene-copy in response to antiviral systems, there is precedence for a copy number variation mechanism used by the virus. Considering that there is no coding element in this region, the most likely need to vary this region is structural; therefore, we predict that the STR region may be under selective pressure to provide stability to this region of the genome.